Amazon Aurora
Amazon Aurora is a MySQL- and PostgreSQL-compatible, fully managed relational database engine on Amazon Web Services, built as part of Amazon RDS to exceed the performance and availability of traditional enterprise databases.
AI Practitioner focus
- Aurora PostgreSQL-Compatible Edition can use the
pgvectorextension to store embeddings and perform similarity search for RAG while retaining relational application data and transactions. - Choose it when vectors and relational SQL belong together; use Amazon OpenSearch Service for search-centric hybrid retrieval and Amazon Neptune for graph-centric relationships.
Key points
- Performance: up to 5x greater than MySQL on RDS, and up to 3x greater than PostgreSQL on RDS.
- Replication: 2 copies of data written in each Availability Zone, across a minimum of 3 AZs — 6 copies total per region — held on a shared storage volume.
- Aurora Replica (in-region) — up to 15 read replicas.
- MySQL Read Replica (cross-region) — up to 5 read replicas; cross-region replicas improve disaster recovery and provide faster reads in regions closer to application users.
- Managed as part of Amazon RDS via the RDS Management Console, CLI commands, and API operations covering patching, backup, recovery, database setup, and failure detection and repair.
- Aurora Serverless — automatically starts up, scales compute capacity based on demand, and shuts down; suited to infrequent, intermittent, or unpredictable workloads with the least management overhead.
- Diagram: one region with AZ 1 / AZ 2 / AZ 3 — reads served from AZ 1 and AZ 3, writes go to AZ 2, all over a shared storage volume holding two copies of data per AZ.